Last week, a piece of news came out that, at first glance, seems like just another corporate move: Meta is studying launching a cloud infrastructure business, internally named Meta Compute, to sell computing power and access to AI models to other companies, including direct competitors. It’s not the kind of headline that becomes group chat material on WhatsApp. But it’s exactly the kind of thing I stop to read twice, because it reveals where the real game is being played.
I started thinking about it because at Superplural we live on both sides of this story. On the one hand, we help brands use AI in practical ways, without hype. On the other, we deal directly with the pain of clients who bought the promise of “digital transformation with AI” and then found, right when it came time to implement it, that the problem was never the model. It was the infrastructure behind it.
Meta entering this game is the clearest sign so far that the fight has moved to a different playing field.
What Meta is actually proposing
According to what was reported in the Social Media Today piece and later confirmed by other outlets, the plan has two fronts. One is selling access to Meta’s own models, such as those in the Muse Spark line, charging based on usage, in a model similar to AWS Bedrock. The other is more audacious: renting raw computing capacity, something that today is the territory of specialized players like CoreWeave and Nebius and, of course, the three major players: AWS, Google Cloud, and Azure.
The reason isn’t a secret. Meta has already committed something between $125 and $145 billion in AI infrastructure just for 2026, for a total that exceeds $600 billion projected over three years. That’s an amount of installed capacity that, as Zuckerberg himself has commented publicly, has people knocking on the door wanting to buy it at a premium. It makes economic sense to sell the surplus. But the data that catches my attention most is another: the move is happening precisely when the market is most skeptical about the returns on these AI investments, and Meta’s offers in this area still haven’t gained traction as expected.
In other words: when the product doesn’t take off the way it was expected, it becomes infrastructure. It’s a classic pivot, just on a scale of hundreds of billions of dollars.
AWS and Google Cloud won’t lose this fight for free
Here’s the point that usually gets left out of these news stories: having spare computing capacity isn’t the same thing as having a cloud business that actually works. AWS and Google Cloud spent more than a decade building what truly underpins an enterprise cloud: dedicated sales teams, compliance certifications, SLA contracts, 24/7 support, integrations with virtually any legacy stack that a large company still carries. None of that gets solved with capex.
Migrating a critical workload from a provider is, in the real-world practice of the companies we serve, one of the most painful and risky processes there is in IT. It’s not like switching out an A4 paper supplier. It involves sensitive data, dependencies nobody documented properly, teams that learned how to operate within a specific ecosystem. That’s why companies are cautious, and why even a new name, with deep pockets like Meta, doesn’t move AWS and Google Cloud overnight. That caution is real, and it’s healthy.
What this teaches people who just wanted to “use AI”
This is where I want to stop talking about Meta and start talking to whoever is reading this and thinking about modernizing their company’s own operations.
There’s widespread confusion in the Brazilian market between “adopting AI” and “having infrastructure ready for AI.” These are completely different things. A company can sign up for ChatGPT Enterprise for the entire team tomorrow and remain just as stuck as it was before, because the bottleneck was never access to the model. It was the data spread across spreadsheets, the legacy system without an API, the process that has no owner, the decision about which cloud, which database, which architecture will support all of that over the next five years.
Weber had a concept I use a lot when I think about companies at this moment: rationalization. The idea that the modern world moves forward by organizing processes in increasingly systematic ways, and that whoever understands this gains a structural advantage over those who only react to trends. The fight among Meta, AWS, Google Cloud, and Azure over who will be the “pipeline” for AI is, at its core, a rationalization race in infrastructure. Whoever wins won’t be the one with the most impressive model in a demo. It will be whoever can package it, charge for it, and sustain it in a predictable way, at scale, for the company that doesn’t want to become anyone’s technical hostage.
And that’s where my healthy skepticism lives regarding part of the AI hype we see pass by every day in the feed. A good company isn’t the one with the newest AI. It’s the one that has a solid foundation built so any AI can run on top of it without everything breaking down in eight months.
The role we play on this playing field
Superplural obviously doesn’t compete with Meta, AWS, or Google Cloud. But we live exactly in the middle of this dispute: it’s us who meet with the client before the decision of which cloud to use, we design the migration process without losing data or operations along the way, and we handle implementation until the system is actually running, not just signed into the contract. It’s behind-the-scenes work, without headline glamour, but it’s what separates a company that “has AI” from a company that truly changed how it operates.
If the race for cloud infrastructure is heating up among giants, the message for the mid-sized and large companies we serve is simple: the time to decide on architecture, migration, and implementation carefully is now, before the rush to not fall behind turns into yet another AI project that never leaves the paper. That is literally Superplural’s specialty: real digital transformation, from the migration process to implementation working in practice.
If your company is at that decision point, you know where to find us.
